Well, first are you conducting separate regressions or multivariate/SEM? If you're using separate regressions you'll already run into issues controlling for interdependence.

Assuming it is a multivariate regression, you can easily check the variance inflation factor (VIF) for each predictor, that will quickly tell you if there is problematic multicollinearity.

When it comes to moderators the purpose of them is to explain how (or under what conditions) a predictor affects a dependent variable. Its utility doesn't hinge solely on its correlation with the DVs but on its theoretical role and the interaction effects it creates. You could also compare "fits" (R2, AICs) with a model without the moderator and assess which is better

Marius Ole Johansen

Thank you very much for your complete answer.

• Having 14 predictors in regression analysis with multiple dependent variables is feasible, but multicollinearity could become a problem depending on how strongly correlated the predictors are with each other. Multicollinearity occurs when two or more predictors are highly correlated, which can make it challenging to determine the individual effect of each predictor on the dependent variable(s). To assess multicollinearity, you can calculate Variance Inflation Factors (VIF).
• Generally: VIF values greater than 10 suggest high multicollinearity. Even moderate correlations can cause issues depending on the model's sensitivity. Regarding having three dependent variables, multivariate regression models such as Multivariate multiple regression can handle multiple dependent variables simultaneously, but it's important to ensure the predictors are not strongly intercorrelated to avoid redundancy. If multicollinearity is identified, potential solutions include:Removing highly correlated predictors. Applying techniques like Principal Component Analysis (PCA) to reduce dimensionality. Using ridge regression or LASSO regression if appropriate.
• A correlation value below 0.6 suggests that the moderator is not strongly correlated with the dependent variables, which generally reduces concerns about collinearity but does not eliminate its role as a potential moderator A good moderator should interact with predictors and explain variance in the dependent variables under certain conditions, even if correlations are modest. Hence, the decision should depend on:Theoretical justification: If the literature or your theoretical framework supports the role of this variable as a moderator, you can proceed. Empirical significance: Even a moderate correlation could allow the moderator to affect relationships, especially if interaction terms in your regression model show significance.

Rajendra K. Gupta

Thank you very much for your time and dedication.